def _build_model(self):
"""
Build the crucial components for model training
"""
if self.is_loadmodel is False:
_config = {
'input_size': 768,
'layer_hidden_sizes': self.layer_hidden_sizes,
'num_layers': self.num_layers,
'bias': self.bias,
'dropout': self.dropout,
'bidirectional': self.bidirectional,
'batch_first': self.batch_first,
'label_size': self.label_size
}
self.predictor = callPredictor(**_config).to(self.device)
self._save_predictor_config(_config)
if self.dataparallal:
self.predictor = torch.nn.DataParallel(self.predictor)
self.criterion = callLoss(task=self.task_type,
loss_name=self.loss_name,
aggregate=self.aggregate)
self.optimizer = self._get_optimizer(self.optimizer_name)
def _build_model(self):
"""
Build the crucial components for model training
"""
if self.is_loadmodel is False:
_config = {
'input_channel': 768,
'nhid': self.nhid,
'n_level': self.n_level,
'kernel_size': self.kernel_size,
'hidden_size': self.hidden_size,
'label_size': self.label_size
}
self.predictor = callPredictor(**_config).to(self.device)
self._save_predictor_config(_config)
if self.dataparallal:
self.predictor = torch.nn.DataParallel(self.predictor)
self.criterion = callLoss(task=self.task_type,
loss_name=self.loss_name,
aggregate=self.aggregate)
self.optimizer = self._get_optimizer(self.optimizer_name)